AI in Construction: A Blueprint for Success

The construction industry is set to see major growth in the use of artificial intelligence (AI) technologies in the coming years. Valued at $530.2 million in 2023, the AI construction market is projected to hit $713.2 million in 2023, with a compound annual growth rate of 34.5% according to a report by Straits Research. AI technologies that enhance performance, efficiency, and safety will help construction companies of all sizes bring the benefits of automation from the factory floor out to real-world job sites.

This article will explore how AI not only improves accuracy in jobs that are tedious, like reading blueprints, but also how AI-enabled robots can complete dangerous tasks on the job site to prevent workers from being injured or killed. In addition, AI helps companies address skilled labor shortages by enabling crews to focus on complex tasks that require human decision-making, effectively getting more done with the potentially limited number of people available.

Inspection and Reading: Faster and More Accurate Results

AI can be used to automate tasks such as inspection and reading where human error is common. Someone needs to count every door, every light switch, every plug cover in a building. Errors in document reading can lead to coming up short — or over purchasing — building materials. Coming up short can cause delays in completing tasks. Over purchasing leads to unnecessarily higher costs.

Consider a recommissioning project where a plant with substantial plumbing needs to have its gaskets serviced. Before AI, people would need to review the existing pipeline and instrumentation diagrams to identify where each of the gaskets is located. This is a tedious, highly detailed process prone to human error. However, missing even a single worn gasket, due to this error, in a complex assembly could have catastrophic results. Thus, multiple people may need to review the diagrams independently to ensure all the relevant gaskets are identified.

“With AI-assisted reading, now you can measure once, so to speak,” says David Golembiewski, a senior account executive at Landing AI, which provides visual AI software-as-a-service (SaaS) capabilities to companies to help them extract value from their unstructured data. Specifically for the construction industry, Landing AI assists builders and site managers in reading blueprints, architectural diagrams, site plans, instrumentation diagrams and other construction documents.

“With AI, we can achieve a 50% savings in time by automating 80–90% of labeling tasks,” says Golembiewski. “By having a computer identify the majority of locations, human reviewers can focus their attention on resolving the questionable locations marked by the AI and get the job done in significantly less time.”

Using AI-assisted tools has a relatively low barrier to entry. First, the relevant documents are made accessible to the computers, either as CAD files or pictures taken of paper documents if computer files are unavailable. A person then spends a short time identifying or “labeling” key features to help train an AI model. For example, if the company wants to track where every filter in an HVAC system is located, the person labels a sample of filters on the documents until the AI model can successfully identify them on its own. The same can be done for doors, gaskets, pipes and other features that need tracking at some point along the construction process. Note that once they’ve been created, companies can use these AI models across jobs that use similar types of documents.

Simplifying Maintenance

AI-based tools can also help builders begin to automate the creation of maintenance plans for new or existing construction. Rather than having maintenance personnel read through hundreds of pages of documents to figure out what they need to do, site managers can now create accurate plans quickly and easily without needing to dedicate staff to the task. 

This approach effectively centralizes information about the construction site. For example, rather than having separate lists or documents that outline the location of the lights or gaskets, this information can be consolidated in a main building plans database. Anyone with access to the building plans can easily find this information when they need it.      

Consider the regular maintenance task of replacing HVAC filters. Today, maintenance personnel need to manually read plans to locate where each filter is, with a chance of missing one or several of them. Instead, an AI system can be trained to extract, detect, and read the data necessary to locate every filter in a building. The building diagrams could be selected to highlight this information, enabling even a person new to the job to find every filter accurately every time. Note that such maintenance plans are dynamic in nature: As changes are made to a building over time, updates to the building plans are automatically reflected in the maintenance plans created by the AI. 

This is just the beginning of how AI can improve construction and maintenance management. According to Golembiewski, soon, AI-assisted inspection and reading will help site managers identify and track build completion progress. A video capture device — either a person with a camera or an autonomous mobile robot (AMR) — will traverse the work site. The captured data will compare building plans with what is currently on the site. This will provide a highly accurate and detailed account of what needs to be done, enabling more efficient allocation and scheduling of construction resources for faster project completion.

Figure 1 shows an example of how visual AI is already being used in extreme environments applications such as undersea inspection. Here, visual AI is used to identify pipe joints and assess their condition. To be accurate, the AI must be able to adapt to the complex and ever-changing floor of the ocean.

Figure 1: Visual AI can be used in extreme environments and applications such as undersea inspection and maintenance. (Source: Landing AI)

Onsite Construction

While increased productivity and efficiency are important, the move towards using robots on work sites for many construction companies has initially been driven primarily by safety concerns. After all, construction can be dangerous. In 2022, 1,069 U.S.-based construction professionals died while working, more than in any other industry sector. This represents 96 fatalities per 100,000 full-time workers or one fatality every 96 minutes. Of these deaths, 38% were caused by falls, slips, or trips.

Figure 2: Robot completing façade application without requiring workers to approach the edge of the building. (Source: Raise Robotics)

Consider tasks such as façade application (see Figure 2), where traditionally, workers have had to work close to the edge or even outside the building to complete the job (click here to see video). Such tasks require safety apparatus that increases cost both in terms of equipment needed and additional time to deploy. With façade application, there’s also a risk of the worker falling, leading to injury or death of the worker and/or those working below. The use of a robot allows for precision work that keeps workers inside the building and eliminates the risk of falling (See Figure 3).

Figure 3: Robots perform precision work while keeping workers safe inside the building and eliminating the risk of falling. (Source: Raise Robotics)

Another important aspect of work safety is repetitive injury. Concrete drilling, for example, takes a toll on a human body day after day and can lead to debilitating conditions like hand-arm vibration syndrome (HAVS).

Raise Robotics is focused on building robots that aid workers by tackling the repetitive and dangerous tasks on commercial jobsites. Consider a typical façade crew of four people. Two workers perform layout and two handle the actual façade work. A robot/human team can perform the work of two people, with the robot performing the dangerous tasks and keeping its human partner safe. Depending on the task and the crew, robots can match or exceed human productivity rates.

Easy to Deploy

Raise Robotics has done substantial development work to make it easier to deploy their robots and can get a robot ready for work on a site in two weeks, depending upon which components are needed for the job and the systems that are in use to manage the job. Gary Chen, CEO of Raise Robotics, says, “One of our advantages is that we don’t need a lot of information from the customer site. Our robots just need a 2D layout such as an AutoCAD drawing or blueprint.” Operators communicate with the robot using a simple graphics interface to identify the work location. The 2D layout has already been labeled with the work points of interest. Once the robot is set up by the operator, the robot deploys vision technology to move and operate within that work site safety as it completes its tasks.

Currently, the ratio is one operator for each robot. “We’re moving towards being able to have one operator for two robots,” Chen says. “We’re also increasing the autonomy of the robots so that once they’ve been set up, the operator can leave a robot to complete the given task while they do their own work.”

The Value of Consistency

One important advantage of using AI-enabled robots is consistency. Site managers are all too familiar that people have productive — and not so productive — days. This inconsistency makes it difficult for site managers to plan effectively, especially when working with multiple, specialized teams. For example, if a drilling crew is working at top efficiency, they may complete a job before the crew following them is ready to step in. Similarly, a crew that is operating below peak or has a member who calls in sick may cause delays that propagate across multiple job sites, leading to complex rescheduling and expensive delays.

Robots offer the advantage of day-to-day consistency. Once set up, they perform their job in a set window of time with reliable efficiency and accuracy site managers can count on. Certainly, robots need time off for maintenance too, but this can be scheduled and planned for so it doesn’t disrupt the master schedule. Robots can also encounter the unexpected, which can cause delays for any crew. However, AI-enabled robots learn over time, in many cases enabling them to handle a wider range of unexpected situations on their own.

Bringing Robots to the Work Site

There are different models for engaging robots on construction jobs. Rather than require its customers to invest in purchasing an entire robot, Raise Robotics leases robots by the month. Leasing gives companies the flexibility to vary their robotic workforce according to current needs. This approach also eliminates high initial investment as a barrier to entry, allowing companies to make small deployment tests to see how well robots work for them.

“At current costs, construction companies will start seeing savings within the first month,” Chen says. “Of course, savings vary based on location and local labor rates. However, it’s not so much about a robot costing less than a person than it is making each person on the job site more effective while keeping them safer.”

Bringing in robots can also help a contractor meet tight deadlines. For example, the speed of a drilling crew depends upon the size of the crew. Several robots can be brought to a job site to help catch up in time and bring the schedule back into line when additional human workers may not be available.       

Part of the engagement process is working with everyone involved, from the contractor down to the crew using the robot, including subcontractors, site manager, and safety teams. “Feedback is a critical part of the deployment cycle for us,” Chen says. “You can’t work out all the issues in the lab. That’s why we go onsite and work with crews to ensure their needs are being met. This is an important part of how we make sure we’re building the right tools for the construction industry." 

Construction robots are connected to the cloud and control software via 5G. This means managers both onsite and offsite have access to operational information. Remote connectivity also makes it possible for Raise Robotics to provide direct, real-time support of robots in case any issues arise. In addition, operational information can also be collected over time to improve AI models as well as to implement advanced features such as predictive maintenance. For example, because robot performance is measured over time, trends in operation can be used to predict when a robot might need maintenance before a breakdown occurs, thus avoiding costly downtime and work disruptions.

Overcoming the Skilled Labor Shortage

Interest in using robots has also risen due to a decline in the availability of skilled labor. “Over the last few years, contractors have been feeling the squeeze from labor shortages,” Chen says. “For this and other reasons like improved safety, we’ve found them to be receptive to considering bringing robots out to the work site.”

Put another way, AI helps augment a person’s capabilities. Assisted by a robot, now one person can do the work of two or more with less risk of injury. With tool changing systems, robots can take over many pneumatic tool-based tasks to allow workers to focus on less physically stressful, higher value-added work.

Golembiewski says, “AI is about improving human capital management. Say you have to dedicate 10 people to looking at instrumentation diagrams and blueprints. Automation of reading these documents augments their ability to perform these tasks more efficiently and accurately. But it also frees them up to do higher value work.”

This illustrates a key aspect of automation: that robots can help people do more whether they are working 10 stories up or sitting in an office with building plans. By automating tedious and repetitive tasks, workers can focus on more complex work that’s more interesting, engaging, and safer.

In Conclusion

The value proposition for AI-assisted inspection and document reading is compelling: using AI-based tools leads to more accurate results that are faster, more efficient, and lower cost than traditional methods. These in turn translate to faster time-to-completion. In addition, higher accuracy means site managers can more precisely purchase building materials as well as more consistently assess how long it will take to complete a job.

The advantages of using robots for repetitive and dangerous tasks are also worthy of notice: greater worker safety, higher efficiency, consistency and a way to extend the capabilities of existing personnel as the labor shortage continues. As Chen says, “AI is truly changing the construction industry for the better.”